Markov Random Field Model-Based Label Classification Method for High-Resolution SAR Image Recovery

In the research of synthetic aperture radar (SAR) imaging technology, there is an increasing interest in effectively using prior knowledge to achieve high-resolution images under downsampling conditions. In order to distinguish the target from the background clutter using prior conditions such as target continuity, this paper reconstructs the SAR image based on the Bayesian maximum posterior method. We construct three hidden variables of the scattering point: intensity, label type and distribution parameters, and then estimate the values of the variables. Among them, we assign a Markov prior to the distribution of label type, and design the energy function of the Markov model to encourage the continuity of labels and distinguish the influence of different neighbors. Simulations validate the proposed method, and the results show that the method can effectively correct the discontinuity of the prior label distribution and eventually iterate to recover a continuous target.